Detecting Signal Spoofing and Jamming Attacks in UAV Networks using a Lightweight IDS

In recent times, security issues pertaining to drones have received a great deal of attention from researchers in networking and communication circles, given their applications in the civilian and defence domains. Object collision avoidance over a trajectory is the only built-in security mechanism in the autopilot system of an unmanned aerial vehicle (UAV). This mechanism, however, cannot protect drones from signal spoofing and hacking attacks. Attacks on UAVs can be triggered from either the ground or flying vehicles in the transmission vicinity medium, by means of which attackers get to control flight operations or manipulate the UAV's autopilot system. An intermittent network connection that disrupts communication in UAVs exacerbates the problem. Hence, a deep learning-based, adaptive Intrusion Detection System is needed for a drone to identify its intruders and ensure its safe return-to-home (RTH). In the proposed IDS, Self-Taught Learning (STL) with a multiclass SVM is used to maintain the high true positive rate of the IDS, even in uncharted territory. A self-healing method in IDS recovery phase uses the Deep-Q Network, a deep reinforcement learning algorithm for dynamic route learning to facilitate the drone's safe return home. Simulation results show the efficiency of the proposed IDS against cyber security attacks on UAVs in terms of accuracy, sensitivity and specificity.